【摘要】 矽肺是我国最为严重的职业病之一,严重危害工人的健康。研究表明,矽肺与粉尘接触量、吸烟量、接尘时间等存在明显的剂量反应关系。基于各矽肺致病影响因子,分别利用指数平滑-神经网络ES-BP(Exponential smoot-hing-BP neural network)和模糊c均值聚类-神经网络FCM-BP(Fuzzy c-means clustering-BP neural network)组合模型对接尘工人未来是否患病以及患病年龄做预测分析。实验结果表明:ES-BP模型能结合原始工人接尘时间队列数据特点,从时间序列上对工人患病年龄进行预测;FCM-BP模型对数据预归类,能极大减小模型复杂度并降低网络训练时间。两种组合模型预测精度均高于BP单独建模预测精度,在工人患病年龄预测中取得了较好的测试效果。更多还原

【Abstract】 Silicosis is one of the most harmful occupational respiratory diseases that are threatening the miners working in dust environment in China deadly.Recently the researchers find that pneumoconiosis obeys the actual dose-response relations with the calendar-year exposure matrix,smoking and individual occupational histories.Two hybrid prediction models based on BP neural network,exponential smoothing and FCM clustering were proposed to predict the possibility and the ages of the miners suffering the silicosis.The experiment results show that the efficiency and the accuracy of the both hybrid models are enhanced greatly compared with the single BP neural network;the BP-ES(Exponential Smoothing)model can make a prediction from the time series datum of dust-exposed workers and the other pathogenic factors;and the model complexity and the training time can be highly reduced with the method of pre-classification in BP-FCM clustering model.The hybrid models are effective methods for silica dust diagnosis prediction.更多还原